Inspiration
AWS announces a great amount of new innovations, services, features, and improvements every year. (Especially after every re:invent!) One can easily lost in tracking these over time. There's gotta be a better way using AI.
What it does
AWS Service Timeline Insights presents a refined view into the chronology of AWS announcements, with keywords spotlighting particular technology or business domains.
How I built it
I played around with the PartyRock. It's super easy to get started. It helped me to kickstart the initial base. I then begin defining the inputs and then tuning the LLM prompts by experimenting with different LLM models.
Challenges we ran into
The top challenge was to find the prompt so the model a) use better data sets while generating answers, and b) answer with accuracy. I solved the data set coverage by specifying the public data sets in the prompt. I improved the accuracy by asking the model to self exam and recover.
Accomplishments that I'm proud of
I built an insight app in a day that would likely take me a couple weeks of work (gathering, parsing, cleaning, preprocessing data, storing data, refining queries). And the most awesome of all, with a new tool, PartyRock.
What I learned
Prompt Engineering requires delicate use of language. While it feels like tuning an AM radio knob sometimes, when it clicks, it's plays beautifully.
What's next for AWS Service Timeline Insights
AWS Service Timeline Insights have following improvements on the roadmap:
- More accurate data. This will likely need additionally trained data sets.
- Predict what will come out as the next feature update for the service.
- Share the view with more folks.
Built With
- partyrock
Log in or sign up for Devpost to join the conversation.